Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
I'm starting this thread, because I've run across what seems like a hundred different performance metrics. Most of them are "ratios" that compare/evaluate various trading aspects.
It would be beneficial for everyone to see what acceptible ratio values are for some of these. I'll start out with a list and you can chime in on what you've observed as bad, good, acceptible, terrible...etc.
P/L factor
Adjust Net Profit Factor
Ulcer Index
Kestner K-Ratio
Semideviation Ratio
R-squared
Return Retracement Ratio
Rina Index
Sharpe Ratio
MaxDD%
Lake Ratio
Sortino Ratio
R-Cubed
Length of Nth longest drawdown
Half of these I've never used....but in my trading, it's hard for me to compare my performance outputs against anything other than my own strategies and reports.
Essentially, if I run a report and the index/ratio gets better (depending on the convention) then I guess that's good. But it's hard to know just what is a "good" or acceptible Sharpe Ratio.
So I'll start....I've been using the RINA Index, which takes the select total net profit, divides it by the average drawdown and divides it again by the percent time in the market.
I've been using RINA because I don't backtest 36 months and so the Sharpe is out for me.
RINA rewards you for the profit you make, with the least amount of time "exposed" in the market. Strategies that have a "get in, get out" will benefit from this ratio.
I've found that a "good" RINA index is one over 100 and an acceptible ratio is between 30-100 and anything less than 30 (even negative) is not so good.
I've also noticed on this board that an unwritten rule or goal is to have an average profit to average loss of at least 2:1. That seems to be a theme among those who make a living.
Examining the 2:1 profit/loss requirement, you get the sense that strategies that feature a mountain of small bites are much more succeptible to changes in the market.
For instance, if you have a strategy that states "90% of the time, with these conditions, it will reach 5 ticks before it reaches 25 ticks against."
In theory, with commission and burden, at a 5/25 (or 1/5 PL ratio), you'd need to win about 6/7 or 86% of the time.
When looking at this closer, if you've observed a 90% win rate and you're averaging a certain yield/time, it only takes a couple of losers in a row to REALLY kick you in the crotch. Essentially, you have to drop from 90% to 86% and now you're under water. Not only that, it only takes 2 or 3 consecutive losers to put you in the hospital.
Conversely, a strategy that features a 2:1 PL ratio, and has a 50% win rate, it takes several losers in a row to create a knockout punch, which usually gives the trader enough time to pause, adjust, etc.
It's a shame this thread didn't take off. Are there any other threads that examine this? I am very interested in objectively measuring my trading account performance. What do hedge funds and prop firms do? I don't want to over-complicate it either...
I am interested in a measure that shows the risk-adjusted performance with no upside penalty (so that excludes Sharpe), and also take into account peak-valley drawdowns. So, lower peak-valley drawdowns as a percent of equity (IOW variance) would be great. I also want my formula to "understand" trading profit withdrawals from the trading account and not interpret them as losses.
I am currently tracking:
Total net profit factor: SUM of All net winning trades / SUM of All net losing trades
R for each trade as defined by Van Tharpe. In other words the Initial Risk/Reward, this is clumsy because it does not take into account the time component. For example if you are in a winning trade for 2 hours, but move to breakeven after 10 minutes, it is much better performance than moving to breakeven (or higher) much later in the trade, after say, an hour, because your account drawdown is protected the sooner you remove your risk.
Average winning trade duration, average losing trade duration: I want to see myself in winning trades many times longer than losing trades (cut losses short, let profits run).
Cumulative net profit: the most important
Max peak-valley drawdown: max anguish sustained
Max consecutive losing trades: not too interested in this, but nice to know
I'm wondering if there's a measure that takes all the above into account and rewarding low downside volatility coupled with high upside volatility. Any ideas?
I have looked into these so far:
Calmar
Stableford Anxiety Index
Sortino Ratio
Sharpe Ratio
Bias Ratio
V2 Ratio
MAR
And have come out a little bit confused!
I started looking into modern portfolio theory as well but it got way too academic and impractical, and I feared I was disappearing down a rabbit hole.
A few of the mentioned variables are related to the initial amount to trade with. So it is important to include some money management related values. Further it is important how many trades are made. To define the drawdown aspect you could also include the ulcer index.
It is a very challenging task to valuate a system with a few variables - but i'm very interested in that, to compare my own systems.
Good points. After struggling with all sorts of metrics I decided to stick with Sharpe ratio, profit factor , number of trades and win rate. Basically when I develop systems I look for those with highest win rate and highest number of trades with Sharpe > 1 and profit factor > 2. Especially the win rate is critical as also argued in this blog because it is inversely proportional to risk of ruin.
I don't want to oversimplify your "issue", but why not calculate all the performance metrics that you know. And then calculate a x-period correlation coefficient to see how well the backtest performance measure correlate over time?
In my view, the higher a backtest performance measure would correlate with itself over time, the more "solid" it is to determine decisions on. That way, you can "trust" that when it signals a low value in period A, it will also give low values in period B, C, and so on.
Vice versa, if you know that a certain measure can be high, low, high, low, low, etc. and not related to net profit, well then it might be just a fancy formula devised by someone to impress others.